Date of Completion

6-5-2020

Embargo Period

6-5-2021

Keywords

Bias Estimation, Sensor Measurements, Target Tracking

Major Advisor

Peter Willett

Co-Major Advisor

Yaakov Bar-Shalom

Associate Advisor

Shengli Zhou

Field of Study

Electrical Engineering

Degree

Doctor of Philosophy

Open Access

Open Access

Abstract

In the field of target tracking there are many challenges associated with error in sensor measurements. Sensors can be subject to sources of consistent error such as temperature warping of the physical equipment, environmental effects in the surrounding atmosphere, and internal clock errors. These errors are known as sensor measurement biases. Biases can corrupt many different types of sensor measurements with many different error models. To overcome such error it is necessary to estimate the biases affecting the system. Two primary methods of bias estimation are examined in this work. The first is simultaneous target state and sensor bias estimation, which assumes a target motion model and applies the biases to it in order to fit the measurement data. The second method is the bias pseudo-measurement method. In this method the sensor measurements are converted into a common Cartesian reference frame and differenced to eliminate the true target state and estimate the biases alone. The advantage of this method is that estimation of the biases is decoupled from the target state estimation. As part of this method it is necessary to use an unbiased conversion to convert the original sensor measurements into Cartesian. This is because of the conversion bias present as a result of the noise passing through the nonlinear conversion from one type of measurement to another. The estimation of the measurement bias can be hampered by the presence of the conversion bias. In this work a variety of methods of bias estimation and unbiased conversion are examined for different types of sensors. Both of the bias estimation methods can be combined with Maximum Likelihood methods. In simulations the bias estimation methods are compared to the Cramer-Rao Lower Bound and each is shown to attain it, meaning that the methods are efficient.

Available for download on Saturday, June 05, 2021

Share

COinS